Codex da-vinci really slow, fails to complete

Is anyone else seeing really slow response times from Codex da-vinci? Noticing issue not just with the api but also in the Playground which throws up a ‘network error’ often.

Model settings below:

Model: code-davinci-002
Temperature: 0
Prompt tokens: 4100
Max length response: 256
Top p: 1
Frequency penalty: 0
Presence penalty: 0
Best of: 1

Been using this for months but started seeing issues on Mar 2 (when there was a broader api outage) and it continues to persist. Can work around latency on application side but in some cases, it is so slow that it seems to give up on the completion, which voids use of the api.

1 Like

Yeah, code-davinci-002 is very slow, almost unbearably slow, timing out.

FWIW, I have been drafting code with the new chat API method and it’s fast enough and works fine.

Example 1

Unformatted Output (Work in Progress)

Example 2

Unformatted Output (Work in Progress)

HTH

:slight_smile:

Thank you, will explore how best to move forward. Hope they restore codex da-vinci functionality soon. But some solace I suppose in knowing it’s not just happening to you :sweat_smile: